Crystalline Growth of Rubrene Film Enhanced by Vertical Ordering in

Mar 7, 2013 - A. K. M. Maidul Islam,. ‡. Yaw-Wen Yang,* ... National Synchrotron Radiation Research Center, Hsinchu, 30076, Taiwan. ‡. Surface Phy...
2 downloads 0 Views 1MB Size
Article pubs.acs.org/Langmuir

Crystalline Growth of Rubrene Film Enhanced by Vertical Ordering in Cadmium Arachidate Multilayer Substrate Chia-Hsin Wang,† A. K. M. Maidul Islam,‡ Yaw-Wen Yang,*,†,§ Tsung-Yu Wu,§ Jian-Wei Lue,§ Chia-Hung Hsu,† Sumona Sinha,‡ and Manabendra Mukherjee‡ †

National Synchrotron Radiation Research Center, Hsinchu, 30076, Taiwan Surface Physics Division, Saha Institute of Nuclear Physics, 1/AF Bidhannagar, Kolkata 700064, India § Department of Chemistry, National Tsing-Hua University, Hsinchu, 30013, Taiwan ‡

S Supporting Information *

ABSTRACT: The growth of highly crystalline rubrene thin films for organic field effect transistor (OFET) application remains a challenge. Here, we report on the vapor-deposited growth of rubrene films on the substrates made of cadmium arachidate (CdA) multilayers deposited onto SiO2/Si(100) via the Langmuir−Blodgett technique. The CdA films, containing 2n+1 layers, with integer n ranging from 0 to 4, are surface-terminated identically by the methyl group but exhibit the thickness-dependent morphology. The morphology and structure of both CdA and rubrene films are characterized by X-ray reflectivity (XRR), X-ray diffraction (XRD), near-edge X-ray absorption fine structure (NEXAFS) spectroscopy, and atomic force microscopy (AFM). Crystalline rubrene films, evidenced by XRD and marked by platelet features in AFM images, become observable when grown onto the CdA layer thicker than 5L. XRD data show that vertical ordering, that is, ordering along surface normal, of CdA multilayer substrates exerts a strong influence in promoting the crystalline growth of rubrene films. This promoted growth is not due to the surface energy of CdA layer but derived from the additional interaction localized between rubrene and CdA island sidewall and presumably strengthened by a close dimensional match between the a-axis of rubrene lattice and the layer spacing of CdA multilayer. The best OFET mobility is recorded for 9L CdA substrate and reaches 6.7 × 10−2 cm2 V−1 s−1, presumably limited by the roughness of the interface between CdA and rubrene films.



INTRODUCTION Organic thin film transistors (OFETs) based on π-conjugated semiconducting materials have received a great deal of interest due to their envisaged applications in low-cost flexible electronics.1−4 Thanks to the rapid advancement made in material synthesis and processing as well as the improved understanding for device physics, nowadays, several types of single-crystal OFETs already yield carrier mobilities in excess of 1 cm2 V−1 s−1, a benchmark value referenced to amorphous silicon. Rubrene (C42H28) transistors exhibit particularly high mobilities ranging from 1 to 18 cm2 V−1 s−1 for single crystal devices.5−9 Unfortunately, the mobility of rubrene thin-film transistor is much lower, ranging from 10−6 to 10−2 cm2 V−1 s−1 because rubrene film tends to grow into amorphous form.10,11 This predominant amorphous growth owes its origin to several factors such as the existence of a myriad of adsorption conformations derived from three-dimensional (3D) geometry of rubrene molecule,12 the need to surmount an energy barrier of 205 meV as the tetracene backbone of rubrene undergoes a change from twisted to planar conformation when forming a solid material from gas phase,13,14 and thermodynamic limitation due to a low substrate working temperature (∼100 °C) that is capped by rubrene desorption from surfaces.15 Consequently, the vacuum-deposited rubrene thin film is © 2013 American Chemical Society

typified by polycrystalline spherulites interspersed with amorphous form.15−17 A readily implemented method to improve the quality of rubrene film is to alter the surface energy of the substrate to promote a better crystalline growth. Different types of surface modifiers including organosilanes,18 organophosphonate,19 and polymers10,20 have been attempted but with limited success, and the corresponding OFET mobilities often fall into the range of 10−2 cm2 V−1 s−1. Meanwhile, several postdeposition treatments aimed at transforming amorphous into crystalline form of rubrene have been reported. Besides a prolonged vacuum annealing method,21 an abrupt annealing of amorphous rubrene film with temperature jumped to 170 °C in 1 min could greatly improve rubrene film, giving rise to an OFET mobility of 1.2 cm2 V−1 s−1.22 The crystalline growth out of the amorphous region of the spherulite can be induced by a very slow substrate-temperature ramping from 80 to 85 °C, yielding an impressive OFET mobility of 2.5 cm2 V−1 s−1.18 Despite this impressive mobility gain, the aforementioned treatments can be quite subtle to perform and hence lack the general applicability. Received: November 26, 2012 Revised: March 3, 2013 Published: March 7, 2013 3957

dx.doi.org/10.1021/la3046912 | Langmuir 2013, 29, 3957−3967

Langmuir

Article

Figure 1. X-ray reflectivity data (○) and fitted curve (thin line) for the CdA films of different thickness grown on SiO2 via Langmuir−Blodgett technique: (A) 1L, (B) 3L, (C) 5L, and (D) 9L. Also shown in the insets are the fitted electron density profiles.

structure and morphology of both rubrene and CdA films are extensively characterized by X-ray reflectivity (XRR), X-ray diffraction (XRD), near-edge X-ray absorption fine structure (NEXAFS), and tapping-mode atomic force microscopy (AFM) techniques. It is found that the platelet-shaped, crystalline rubrene film can be readily formed on thick CdA films. More interestingly, the CdA film with a better ordering along surface normal is found to support a better crystalline growth of rubrene, as evidenced by a strong correlation between out-of-plane XRD intensities of CdA and rubrene films. This intensity correlation, counter to the prediction based on surface energy consideration, can be rationalized by the occurrence of a strong interaction between rubrene and the sidewall of CdA island, facilitated by a close dimensional match between the a-axis of rubrene lattice and the layer spacing of CdA multilayer. The postulated strong interaction helps realize the desired aligned growth of rubrene, with the a-axis of the lattice lined up with the surface normal. Moreover, the mobility of OFET fabricated from the as-produced rubrene films sees an increase from 4.8 × 10−4 to 6.7 × 10−2 cm2 V−1 s−1, as rubrene crystallinity is improved.

Suffice it to say that an effective substrate modification scheme capable of promoting crystalline growth remains to be developed. Beyond the surface modification strategy, several hints of potentially useful role played by the vertical structure of the substrate in affecting the crystalline growth have been reported. Li et al.19 found an enhanced crystallization of rubrene film on 6,13-pentaenequinone (PQ)/octadecylphosphonic acid (ODPA)/SiO2 and claimed that a bilayer-ODPA played a key role in adjusting the morphology of thin insulating PQ layer, which in turn promotes the formation of crystalline rubrene film. The rubrene OFET fabricated this way had a rather high mobility of 0.35 cm2 V−1 s−1. Similar beneficial effect of the bilayer ODPA in promoting the rubrene crystallization was also noted when PQ was replaced with 6,13-diazapentacene (DAP), and the resultant OFET mobility reached an even higher value of 0.68 cm2 V−1s −1.23 Mannsfeld et al.24 showed that rubrene, pentacene, and C60 could each nucleate at the bases of rough, tall octadecyltriethoxysilane (OTES) pillars and grew into single crystals in the sealed tubes placed inside a hot furnace. The resultant single-crystal rubrene OFET possessed a mobility as high as 2.4 cm2 V−1 s−1.25 All of these observations prompt us to examine the possibility of whether a substrate exhibiting a purposely built 3D structure can be beneficial to the crystalline growth of rubrene. Langmuir−Blodgett (LB) technique is known to produce thin films of well-defined structural architecture, offering a convenient approach of building up molecular device in succession.26,27 In relevance to OFET application, LB method has been used to produce dielectric layers for pentacene growth28 as well as active layers of poly(3-hexylthiophene)29 and copper phthalocyanine30 on SiO2 substrates. In the present study, we investigate the rubrene growth on LB films of cadmium arachidate (CdA) prepared on SiO2 substrates with the CdA films composed of 2n+1 layers (L), with n from 0 to 4 inclusive. As-prepared CdA layers are all terminated by the same methyl group but possess different morphology. The



EXPERIMENTAL SECTION

LB films of cadmium arachidate (CdA) were prepared using a KSV 5000 instrument. Arachidic acid (CH3(CH2)18COOH, Sigma, 99%) monolayer was spread from a chloroform (Aldrich, 99%) solution in a Langmuir trough on Milli-Q water (resistivity 18.2 MΩ cm) containing cadmium chloride (Loba Chemie, 99%). For LB deposition, the pH of the water subphase containing 5 × 10−4 M CdCl2 was maintained at 6.0 ± 0.2 by sodium bicarbonate (Merck, 98%) at room temperature. After preparation, the monolayer was allowed to equilibrate with the subphase prior to compression for about 10 min. Subsequently, the monolayer was slowly compressed with a constant barrier speed of 3 mm/min until the surface pressure needed for the transfer was reached. The monolayer was kept at a constant pressure for 5 min to facilitate the structural relaxation and then transferred onto a substrate at a transfer speed of 1.5 mm/min. The Si(100) wafers covered with 300 nm thermally grown SiO2 were used for growing LB films and later fabrication of OFETs with rubrene 3958

dx.doi.org/10.1021/la3046912 | Langmuir 2013, 29, 3957−3967

Langmuir

Article

Figure 2. AFM images of various thin films used to support the rubrene growth: (A) OTS/SiO2, (B) 1L CdA/SiO2, (C) 5L CdA/SiO2, (D) 7L CdA/SiO2, and (E) 9L CdA/SiO2. The size of all of the images is 2 μm × 2 μm, except the one in (A) that is measured 5 μm × 5 μm. Also shown beneath each AFM image is the height variation along the designated line in the image.

oscillations are seen for the films thicker than 3L, suggesting a very good quality of the films. The reflectivity curves for the 3L to 9 L are similar to the data reported earlier,26,33 with the bilayer repeat unit made out of two arachidate units bonded to a single Cd cation in a head-to-head manner. As the pH of the water subphase in a LB trough controls the degree of dissociation of arachidic acid, the structure of the interface between SiO2 substrate and CdA film thus varies with the pH value accordingly. For the pH value of 6 used here, acidic amphiphile completely dissociates, and hence the interface is composed of cadmium salt with one cadmium cation associated with two neighboring arachidate molecules.26,34,35 To obtain the information about the thickness and electron density profile of the films, the reflectivity data were analyzed using Parrat formalism36 modified to include interfacial roughness. In brief, the input electron density profiles were divided into several boxes with the thickness of no less than the depth resolution of 2π/qmax = ∼5 Å, where qmax is the maximum wave-vector transferred normal to the surface, and interfacial roughness was kept within 2−8 Å. During the analysis, the roughness of the top surface, electron density, the thickness of the films, and the roughness of the substrate were varied as fitting parameters. The bilayer thickness obtained from fitting the X-ray reflectivity data is 5.534 nm, in full agreement with the earlier data.31,37 Figure 2 shows tapping-mode AFM images of OTS/SiO2 and CdA/SiO2 substrates that are used to support the rubrene growth. The OTS thin film (Figure 2A) is smooth and possesses a random height variation characterized by a rootmean-square roughness (Rrms) of 0.26 nm. For 1L CdA (Figure 2B), the Rrms of the film increases slightly to 0.34 nm, and the smooth surface is marked with a small number of irregular, patchy features with a height jump of ∼0.6 nm. For 5L CdA (Figure 2C), its AFM image is different from those of the thin layers, and the surface is characterized by labyrinths of high regions separated by lower regions with a height difference of 2−4 nm, yielding an increasing Rrms of 1.1 nm. The AFM image for 7L CdA (Figure 2D) is dominated by regularly spaced, slender features that take up a height of 4−10 nm and a width of about 50 nm, with the Rrms of the whole image worsened to 2.2 nm. As the thickness of CdA is increased to 9L (Figure 2E), a dramatic change of morphology is noted. The jaggy feature

as an active layer. The wafers were made hydrophilic according to the RCA cleaning procedure by boiling the substrate in a mixture solution of ammonium hydroxide (Merck, 98%), hydrogen peroxide (Merck, 98%), and Milli-Q water (H2O/NH4OH/H2O2 = 2:1:1 by volume) for 10 min. Immediately after a final rinsing with water, the Si wafers were placed into the trough and the LB deposition was started. LB films of different layer thickness (1, 3, 5, 7, and 9 layers, L) were prepared at a surface pressure of 30 mN/m, and the completed LB films were stored in a dry-nitrogen box for later use. Rubrene (Acros, 99%) films were vacuum-deposited onto different CdA layers at 358 K with the flux rate maintained at 2.4 nm/min. The same thickness of 250 nm was chosen on the basis of our observation that rubrene OFET began to exhibit appreciable mobility at this thickness of active layer. The structural characteristic of CdA multilayer is expected to remain the same despite the high substrate temperature in use here based on the previous X-ray diffraction studies in which CdA multilayer was found to dewet from SiO2 at 383 K and proceed to desorb at temperatures higher than 433 K.31,32 The fabrication of top-contact, bottom-gate OFETs was complete by forming Au electrodes on top of the rubrene films via vacuumdepositing gold through a shadow mask patterned with a fixed channel length and width of 50 and 500 μm, respectively. The output and transfer characteristics of the OFETs were measured in ambient condition using a Keithley 2636A dual-channel source-meter instrument. The film topography was characterized by an atomic force microscope (AFM) operating in tapping mode, using either Innova by Veeco or Multiview 1000 by Nanonics Imaging Limited. X-ray diffraction (XRD) study was carried out using MAC MXP-18 and BRUKER AXS D8 ADVANCE diffractometer with Cu Kα radiation (λ = 0.154 nm). Near-edge X-ray absorption fine structure (NEXAFS) spectroscopy was performed at Wide-Range beamline (BL24A) of NSRRC based on total-electron yield (TEY) and Auger electron yield (AEY) methods with Auger electrons detected by a SPEC PHOBIS 150 electron energy analyzer. X-ray reflectivity (XRR) measurements were performed at Saha institute using an X-ray diffractometer (D8 Discover, Bruker AXS) with Cu Kα radiation.



RESULTS Figure 1 shows the oscillation of X-ray reflectivity as a function of momentum transfer qz for the different CdA films examined here, where qz is equal to 4π sin θ/λ with θ corresponding to the incident angle (=exit angle) of X-ray with a wavelength of λ. Also shown in the inset is the fitted electron density distribution function across film depth. Pronounced Bragg 3959

dx.doi.org/10.1021/la3046912 | Langmuir 2013, 29, 3957−3967

Langmuir

Article

Figure 3. AFM images of rubrene films grown on various substrates: (A) OTS/SiO2, (B) 1L CdA/SiO2, (C) 3L CdA/SiO2, (D) 5L CdA/SiO2, (E) 7L CdA/SiO2, and (F) 9L CdA/SiO2. The thickness of rubrene films is 250 nm for (B)−(F) and 500 nm for (A). The image size is 10 μm × 10 μm in (A), and 5 μm × 5 μm in (B)−(F).

characteristic of 5L and 7L CdA substrates is replaced by wide, flatter terrains separated by a height difference in the multiple of ∼5 nm, leading to an increase of Rrms to 2.95 nm. As noted earlier, the present CdA films were prepared at a pH of 6 to ensure a complete dissociation of arachidic acid so that the LB films were made up of the salt form of cadmium arachidate without incorporating arachidic acid. Therefore, the images in Figure 2C and D are similar to the so-called “skeletonized” CdA multilayers that have their trapped arachidic acid molecules completely removed by benzene soaking.35 Figure 3 shows the tapping-mode AFM images for 250 nmthick rubrene films grown on various CdA layers. Also shown for comparison is the image of a 500 nm-thick rubrene film grown on OTS (Figure 3A). This surface is filled with deeply ridged features running loosely parallel to each other with a feature width of 0.1−0.2 μm, similar to the earlier observation.17 For the rubrene film deposited on 1L CdA, the morphology changes dramatically to bun-shape features exhibiting a diameter of ∼1 μm and a height of about several hundred nanometers. It is worth noting this unexpected morphology change because OTS and 1L CdA substrates are structurally analogous, same surface termination (−CH3), similar molecular dimension (C18SiO− vs C19COO−), similar molecular packing, and similar surface roughness for the resultant films (see Figure 6). For 3L CdA, the rubrene morphology is dominated again by segmented features found for OTS substrate (Figure 3C). Only when CdA film becomes thicker than 5L does the morphology of overgrown rubrene films become unchanged and exhibit characteristic micrometersized platelet (Figure 3D−F). It is surprising to observe that the morphology of the rubrene film changes with the layer thickness of CdA because all of the CdA layers are terminated by the same methyl groups and differ only in the way of how the CdA bilayers are stacked together in forming the films. This unexpected finding seems to suggest that substrate-surface property cannot be solely responsible for the change of the rubrene growth behavior. Figure 4 shows out-of-plane X-ray diffraction patterns obtained for 250 nm-thick rubrene films grown on five CdA/

Figure 4. θ−2θ XRD patterns for 250 nm-thick rubrene films grown on CdA/SiO2 substrates of different CdA layer thickness: (A) 1L, (B) 3L, (C) 5L, (D) 7L, and (E) 9L. The (h00) index designates the order of diffraction from the rubrene films, while the other nonlabeled diffraction peaks are due to the CdA multilayer with a d-spacing equal to a bilayer thickness of 5.538 nm. Note the peak appearing at 2θ = 4.78° corresponds to the third-order diffraction.

SiO2 layers. The observed peaks are all derived from rubrene and CdA films, with those labeled with (h00) indices referred to rubrene. The assignment of rubrene diffraction peaks is consistent with a rubrene crystallite possessing orthorhombic structure38 (a = 26.86 Å, b = 7.193 Å, c = 14.433 Å, α = β = γ = 90°) and orienting itself with its b, c axes of the unit cell falling on the substrate surface. Major diffraction peaks observed at 2θ = 6.58°, 19.82°, and 40.26° (2θ range truncated for clarity) are consistent with earlier reports,39,40 and assigned to (200), 3960

dx.doi.org/10.1021/la3046912 | Langmuir 2013, 29, 3957−3967

Langmuir

Article

(600), and (1200), respectively. Another commonly observed peak indexed by (002) appearing at 2θ = 12.26° is not observed here. This peak is derived from a different growth of rubrene crystallite with its a, b axes aligned with the substrate surface. All of the other peaks not related to rubrene film are attributed entirely to a single CdA structure that has a d-spacing of 5.538 nm, and the obtained CdA diffraction patterns agree with the data reported earlier.41 Moreover, the CdA diffraction patterns for rubrene on CdA and pure CdA films are the same because the thickness of rubrene overlayer (250 nm) is much smaller than the calculated X-ray attenuation length of 0.19 cm at Cu Kα X-ray energy. The intense peak at 2θ = 4.78° is due to the third-order diffraction. The d-spacing obtained from XRD data agrees completely with the oscillation period of 5.534 nm inferred from XRR data fitting. Furthermore, the intensities of CdA diffraction peaks increase with increasing CdA layer thickness as expected, but only up to 7L, and beyond that the CdA diffraction intensity decreases, suggesting a structure reorganization of 9L CdA. Interestingly, the diffraction intensities of rubrene and CdA seem to go hand in hand; that is to say, the CdA film exhibiting stronger diffraction intensity also results in stronger rubrene diffraction. This surprising observation suggests that a better ordering of CdA substrate along the surface normal helps guide the crystalline growth of rubrene overlayer, necessitating a detailed profiling of CdA films along surface normal direction (vide infra). As rubrene film usually contains spherulites in which polycrystalline form is interspersed with amorphous form, Xray diffraction, being sensitive to long-range structural ordering, may not yield readily useful information due to a lack of diffraction features. In comparison, NEXAFS spectroscopy can offer a remedy by providing an averaged tilt angle of rubrene molecule, specifically the tilt angle of the aromatic plane of tetracene core measured from surface normal, θ, irrespective of whether rubrene film is crystalline or not. The θ angle provides a quick estimate for the propensity of the molecular orientation ordering in the films. For the rubrene crystallite producing aforementioned (h00)-indexed diffraction, the tetracene plane is perpendicular to the substrate surface and θ is equal to 0°. Thus, any nonzero tilt angle can be contrived as evidence for the presence of other rubrene conformations. The polarizationdependent NEXAFS spectra for 250 nm-thick rubrene film grown on 7L CdA, the most crystalline substrate, are presented in Figure S1. The θ angle can be determined by curve-fitting the variation of α-resonance intensity, also obtained by fitting, with X-ray incidence angles, as reported before.42 Tetracene plane of rubrene is found to tilt at 22° from the surface normal. This deviation of tilt angle is explained by the coexistence of amorphous and crystallite forms in rubrene films and/or an alteration of α-peak intensity ratio due to a possible rubrene oxidation43 incurred during air-exposed sample transfer between two chambers. The exact cause for this deviation is not known at present. Figure 5 presents two plots derived from transfer characteristic measurement of the OFETs operating in saturation mode using different rubrene films but of the same thickness of 250 nm as the active layers. The output characteristics were obtained first to validate the OFET behavior of the devices and identify the saturation region. Field-effect mobility (μ) is extracted from the saturation regime using the relationship: IDS =

W C tμ(VGS − VT)2 2L

Figure 5. Transfer characteristics of OFETs operated in saturation region with the active materials of 250 nm-thick rubrene grown on various CdA/SiO2 substrates. (A) log |IDS| versus VGS plot, and (B) | IDS|1/2 versus VGS plot. VDS fixed at −25 V.

where IDS is source-drain saturation current, W and L are channel width and length, respectively, Ct is total capacitance per unit area for gate dielectric that includes CdA, VGS is gate voltage relative to source electrode, and VT is threshold voltage. The mobility and threshold voltage are proportional to the slope and x intercept in the plot of |IDS|1/2 versus VGS, respectively. Figure 5A shows the variation of log|IDS| versus VGS, while the |IDS|1/2 versus VGS plot is shown in Figure 5B. For the two OFETs fabricated on 1L and 3L CdA/SiO2, the drain current is rather low, yielding a meager mobility in the 10−4 range. As CdA thickness is increased above 5L, the drain current begins to see a large increase. The highest hole mobility of 6.7 × 10−2 is recorded for 9L CdA, not 7L CdA, the most crystalline film. The electrical parameters of the devices including current on/off ratio are summarized in Table 1. In evaluating the total capacitance of dielectrics, Ct, the contribution from CdA layer needs to be considered as well. A SiO2 surface with a CdA film situating on top can be regarded as two capacitors stacked in series, and the applicable equation is 1/Ct = 1/CSiO2 + 1/CCdA. Monolayer film like 1L CdA makes a negligible contribution to the total capacitance because of its relatively large capacitance associated with small layer thickness. Yet for the thickest film like 9L CdA, its capacitance drops to a value comparable to that of 300 nm-thick SiO2 (=10.8 nF/ cm2), necessitating a correction to the total capacitance. The capacitance of 9L CdA film is calculated to be 85 nF/cm2, using the reported total dielectric constant of 2.444 and a thickness of 24.9 nm, assuming a homogeneous layer. Because of this capacitance decrease, a 12% increase to the mobility should be adjusted for the 9L CdA case.



DISCUSSION The indication by the XRD data that the CdA substrate with a better ordering along surface normal is conducive to a better crystalline growth of rubrene overlayer is somewhat surprising

(1) 3961

dx.doi.org/10.1021/la3046912 | Langmuir 2013, 29, 3957−3967

Langmuir

Article

Table 1. Electrical Parameters Obtained from OFETs Fabricated with 250 nm-Thick Rubrene Grown on Different Cadmium Arachidate Layers on SiO2 mobility (cm2 V−1 s−1) substrate 1L 3L 5L 7L 9L

CdA/SiO2 CdA/SiO2 CdA/SiO2 CdA/SiO2 CdA/SiO2

2 a

cal. capacitance (nF/cm ) 10.6 10.4 10.1 9.8 9.6

avg 3.7 3.9 1.8 1.6 3.3

× × × × ×

max

10−4 10−4 10−2 10−2 10−2

4.8 4.3 3.4 2.0 6.7

× × × × ×

10−4 10−4 10−2 10−2 10−2

threshold voltage (VT)

current on/off ratio

−9.3 −7.3 3.2 1.0 1.6

104 104 104 106 106

a

Substrate capacitance was calculated by assuming a homogeneous CdA layer with the dielectric constant of CdA set equal to 2.4. Also, the capacitance of 300 nm-thick SiO2 was 10.8 nF/cm2.

for the following reason. For a planar substrate like SiO2, or a rough, open substrate with negligible sidewall−adsorbate interaction, the ordering along the substrate surface normal should have nothing to do with the overlayer growth because the overlayer does not interact with the subsurface layers of the substrate. Therefore, the observed improved crystallinity of rubrene film on better-ordered CdA layer suggests that an operative side-to-side interaction between CdA island and rubrene molecule exists and helps achieve the desired aligned growth of rubrene. As a result, a dissection of the CdA films to reveal the sidewall distribution of CdA islands should be a useful undertaking, which forms the first part of this section. Next, X-ray diffraction intensities of the rubrene and CdA films will be compared quantitatively so as to put the intensity correlation on a firmer footing. Taking together these results, a plausible explanation for the enhanced crystalline growth of rubrene film by ordered CdA layer will be given. Finally, the potential drawback in the performance of OFET fabricated with active layer grown on LB films will be discussed. Profiling of CdA Layers. Shown in the right panel of Figure 6 are the three-dimensional (3D) perspective views of the AFM images (Figure 2) for 5L, 7L, and 9L CdA films. The morphologies of 5L and 7L CdA films are dominated by distinct conical-shape structures of different height, dependent on the film thickness. However, for the 9L CdA film, the conical structures coalesce and agglomerate into rolling hills. The height distribution of these features can be obtained through a histogram analysis of the AFM images. Shown in the left panel of Figure 6 are the histograms derived from the analysis of the AFM images presented in Figure 2. The histogram summarizes how the frequency of observing a given height, z, varies with the height itself, with the event frequency normalized to the total number of pixel and expressed in percentage. For both OTS and 1L CdA films (Figure 2A and B), a narrow height distribution is observed with the most probable height located at 1.57 and 2.43 nm, respectively. The 0.86 nm height difference is attributed to the different tilt angle of the hydrocarbon chains, 0° for CdA45 and ∼10° for OTS,45 and the inclusion of cadmium ion in the CdA. It is noted that the bottom most region, or the zero height, may refer to the substrate or overlayer, depending on how the CdA layers are stacked. As the CdA films become thicker, their height distributions become broader. Nonetheless, the maximum height, zmax, marked by the right vertical line underneath each curve in Figure 6, continuously shifts toward higher values as the CdA layer is made thicker. The zmax is located at 3.5, 8.6, 13.9, and 18.9 nm for 1L, 5L, 7L, and 9L CdA films, respectively. The difference in zmax between two successively thicker CdA layers is about 5.3 nm, consistent with a bilayer height of 5.5 nm.

Figure 6. (Left panel) Histograms derived from the analysis of the substrate AFM images shown in Figure 2: (A) OTS/SiO2, (B) 1L CdA, (C) 5L CdA, (D) 7L CdA, and (E) 9L CdA. The histogram displaying the frequency of observing a given height z in the whole image is normalized to the total number of pixel and expressed in percentage. The bin size chosen for frequency counting is 0.1 nm, and the resultant curves are offset for clarity, with their baselines indicated by thin lines. (Right panel) 3D perspective view showing the morphology of 9L CdA (top), 7L CdA (middle), and 5L CdA (bottom) films based on their AFM images.

However, the minimum height, zmin, remains at nearly the same 3.4 nm for 5L, 7L, and 9L of CdA, suggesting an incomplete filling that leaves the bottom-most region unchanged. The difference between zmax and zmin is equal to 5.4, 10.3, and 15.2 nm for 5L, 7L, and 9L of CdA, respectively, roughly equivalent to the integer multiples of the bilayer thickness. It is clear that, during LB fabrication of odd-numbered CdA films, a bilayer is indeed the basic building block, but the block-filling is nowhere near ideal and worsens for the increasingly thicker CdA films. Ordering of CdA along Surface Normal Promotes Rubrene Crystalline Growth. Figure 7 presents a plot of how the total diffraction intensities of rubrene films and CdA substrates are compared to each other based on the data in Figure 4. The total diffraction intensity for each rubrene film was obtained by summing the integrated areas of the diffraction peaks from (200) to (1200) but neglecting weak (1000) because of its overlap with strong silicon wafer peak. Similarly, the total diffraction intensity for each CdA film was obtained by 3962

dx.doi.org/10.1021/la3046912 | Langmuir 2013, 29, 3957−3967

Langmuir

Article

energy of polymer dielectric on the pentacene growth by using poly(imide-siloxane) whose surface energy could be varied between 30 and 48 mN m−1 while the roughness was kept at the same.47 They observed Stranski−Krastanov growth of pentacene on high surface-energy surface, producing large and dendritic grain (∼1.3 μm) and the 3D island growth with small grains (∼300 nm) on low surface-energy surface, in agreement with the prediction based on thermodynamics. However, the mobility of pentacene OFET based on the low surface-energy dielectric is larger by a factor of 5 than pentacene OFET on high surface-energy dielectric, contrary to the expectation that a larger-crystallite OFET should yield higher mobility. This unexpected finding was explained by the better interconnection and tighter packing afforded by small pentacene grains. In contrast, larger pentacene grains formed on the high surfaceenergy dielectric leave voids in the first layer and incomplete layers over the first, resulting in a degraded carrier transport. Capitalizing on the understanding of how the pentacene growth is altered by the surface energy of the substrate, we will show in the following that the observed change of rubrene growth mode is not due to the changing surface energy associated with the thickness change of CdA. To begin with, the SiO2 surface energy has been reported to be ranging from 50 to 65 mN m−1,28,48,49 with discrepancy probably related to the presence of varied amount of surface silanols. After the silanation with OTS, the surface energy of modified SiO2 drops to about 20 mN m−1.49,50 Previous theoretical calculations have placed the free energy of the most stable face of orthorhombic pentacene crystal, that is, (001), at 38 or 50 mN m−1, depending on the structure model and calculation method.51,52 Meanwhile, experimental surface energy value ranges from 34 to 47 mN m−1,53,54 perhaps with the contribution from higherenergy faces partially included. In relevance to the use of LB films as substrates, Nayak et al. reported how the surface energies of different CdA films affect the growth of pentacene and the performance of pentacene OFET.28 Their measured surface energies for 1L, 3L, and 9L CdA films are 49, 36, and 28 mN m−1, respectively. The resultant pentacene crystallites also decrease clearly from an averaged grain size of 0.6 to 0.12 μm as observed by AFM, as the surface energy is decreased, in agreement with the expectation based on the studied by Yang et al.47 The decrease of grain size is corroborated with the observed decrease of out-of-plane XRD intensity. Crystalline rubrene film possesses a surface energy of 33.9 mN m−1,48 similar to that of pentacene (001) face. Comparing the rubrene films grown on OTS and on 1L CdA (Figure 3A and B), one observes much larger features for rubrene on 1L CdA, consistent with the prediction based on surface energy difference (20 vs 49 mN m−1, with water contact angle equal to 108° and 56°, respectively) despite a high structuralsimilarity between two substrate surfaces (Figure 6). However, the prediction of rubrene crystallinity for thicker CdA cases based on surface energy consideration immediately runs into difficulty. The thicker CdA film, with reduced surface energy, is expected to support the growth of smaller-grain rubrene crystallites, yielding weaker X-ray diffraction intensity. Yet, this is opposite to what has been observed, indicating that substratesurface energy cannot be the dominating factor in determining the rubrene growth behavior for thicker CdA cases. Moreover, if surface physical property of CdA were of critical importance in determining the rubrene growth mode, one would not expect a major change of rubrene morphology with the CdA film thickness because all CdA films are surface-terminated

Figure 7. Diffraction-intensity correlation between rubrene overlayer and CdA substrate for all five films based on the data in Figure 4. The line representing a linear least-squares fit to the data is meant to guide the eyes.

summing the areas of peaks associated with diffraction order from 3 (2θ = 4.78°) to 15 (2θ = 24.09°). Because of the overlap between the intense rubrene (200) peak at 6.58° and moderate CdA (400) peak at 6.38°, a two-Gaussian peak fit was done to restore the individual intensity. From the plot in Figure 7, it is clear that the intensities of CdA and rubrene are highly correlated with each other, leading to a perplexing yet interesting conclusion that a substrate with better ordering along the surface normal helps produce a better ordered overlayer. The best ordering for the substrate is found for 7L of CdA, as is the maximum diffraction intensity for the rubrene film. The decreasing diffraction intensity for 9L CdA case is verified by repeating with different samples, and the reduced diffraction intensity is corroborated with the wider height spread found for 9L CdA film in AFM histogram analysis. It is to be emphasized that the effective working range of CdA layer in guiding the ordered growth of rubrene film is limited to the initial-growth stage because rubrene overlayer is much thicker than the CdA. Sidewall Interaction Is More Important than Surface Energy in Governing Rubrene Crystalline Growth on Thicker CdA. From the equilibrium thermodynamics consideration, the layer-by-layer (also called Frank−van der Merwe) growth will be energetically driven if the sum of surface energy of overlayer (σover) and surface energy of the interface between substrate and overlayer (σinter) is smaller than the surface energy of substrate (σsub), that is, σover + σinter < σsub. Conversely, a 3D island growth (Volmer−Weber) will result if σover + σinter > σsub. Between these two extremes, a layer growth for the first few layers and then followed by a 3D growth is possible, and this is termed Stranski−Krastanov growth. Many approaches based on altering the surface energy of dielectric substrates have been devised to effect the desired layer growth of organic films in the hope of fabricating high performance OFETs. There exist many reports, albeit with conflicting conclusion, as to whether a high surface energy substrate should be the choice for better OFET.46,47 The confusion stems from a lack of clear separation among the effects of surface energy, surface roughness, varied interfacial interaction arising from different surface termini, etc. In a well thought-out experiment, Yang et al. studied the effect of surface 3963

dx.doi.org/10.1021/la3046912 | Langmuir 2013, 29, 3957−3967

Langmuir

Article

of rubrene islands at OTS pillars24 and the crystalline rubrene growth promoted by ODPA bilayer19 both suggest the importance of the third dimension, that is, along surface normal, in guiding the crystalline growth of rubrene. These previous observations seem to be in the same vein as the present finding of enhanced rubrene crystallinity by more vertically ordered CdA. A partial support for the aligned growth by CdA sidewall is lent by the observation that rubrene grain seems to selectively grow out of CdA island void. Figure 8 presents AFM images of 5 nm-thick rubrene films grown on 5L CdA layer and OTS, respectively. Rubrene grains are particularly discernible in the phase image due to the enhanced image contrast associated with grain edge. It is found that rubrene does not grow on the high, flat plateau of CdA islands, as evidenced by the conspicuous absence of rubrene grains in the central region of the image (Figure 8B). The effort of identifying how the rubrene grains are spatially registered to the underneath CdA is somewhat hampered by the fact that most of the rubrene grains are too large and hence obscure the underneath CdA features. Nonetheless, several smaller rubrene grains (marked with arrows in Figure 8A) are found to neighbor the void regions of CdA islands, suggesting that rubrene grains germinate from the voids. Figure 8C shows the morphology of 5 nm-thick rubrene film grown on OTS at 358 K. A uniform distribution of rubrene grains that are mostly in amorphous form17 is observed, suggesting the occurrence of random nucleation. Again, surface property of substrate alone cannot account for the growth mode difference observed for 5 nm-thick rubrene film because both 5L CdA and OTS are terminated identically by methyl group. On the basis of the results obtained for thin and thick rubrene films, one can be more assured that the interaction between rubrene and the sidewall of CdA island plays a critical role in governing the rubrene growth. It is noted that the AFM images say nothing about whether rubrene grains are crystalline or not, and if so how they are aligned. A definitive test of Scheme 1 needs to await the result from a vacuum-based, molecularly resolved imaging experiment that is capable of discerning different conformations of rubrene that is free from being oxidized. Adverse Effect of Rough Dielectric Film on Hole Transport of Rubrene. As more ordered CdA substrate supports better crystalline growth of rubrene, it seems reasonable to expect that the hole mobility of rubrene OFET will exhibit the same dependence on the thickness of CdA layer as X-ray diffraction intensity of rubrene film does. Indeed, rubrene OFET mobility generally increases with the thickness of CdA layer, but the highest mobility is found not in 7L CdA, but in 9L CdA case. Of course, the mobility is intimately related to how rubrene grains are distributed and how they are interconnected, particularly for those near the interface with the CdA layers because they are primarily responsible for charge conduction. Unfortunately, these interfacial rubrene crystallites are not amenable to a direct investigation and remain largely unknown. However, there are some specific points needed to be considered when using CdA layer as substrate and dielectric as well. The CdA film with a dielectric constant of 2.4 is also a capacitor capable of storing induced changes. In contrast to SiO2 surface that can be made atomically flat, the top surface of CdA is not flat at all as revealed by AFM images (Figure 2). The topographically rough CdA capacitor presents a problem for effective charge transport among the rubrene molecules riding on top of CdA because neighboring rubrene molecules

identically by methyl group. Again, this is contrary to the observation. Therefore, it is clear that one needs to extend beyond the surface property consideration and recognize the role of the sidewall of CdA island in affecting the rubrene growth mode. The postulated aligned growth of rubrene by the sidewall of CdA islands is depicted in Scheme 1. The matching of Scheme 1. A Schematic Illustrating How the Conformal Crystalline Growth of Rubrene Is Enhanced by Cadmium Arachidate Multilayer via a Matching of Molecular Dimensionsa

a

The hydrocarbon chains of CdA are drawn perpendicular to the surface based on previous studies, and the first layer is depicted in salt form, as described in the text.

molecular dimension between rubrene and CdA is surprisingly good. The thickness of one CdA layer including Cd ion is equal to 2.767 nm (one-half the bilayer thickness), and the a-axis of rubrene orthorhombic unit cell, containing two rubrene molecules, is 2.686 nm. These two numbers are matched to within 3%. Moreover, the carboxylate functional group of CdA might offer a stronger interaction with phenyl groups of rubrene via π−π interaction, further strengthening the effect of aligned growth of rubrene crystal. Once the aligned growth is set in, rubrene molecules will grow into a crystallite with its aaxis of unit cell lined up with surface normal, like the CdA island sidewall does. Afterward, the continued growth into thicker film will become less demanding of correct molecular configuration. Therefore, the thicker CdA film, having more CdA structure units available to guide the aligned growth of rubrene over a longer distance, will result in the growth of larger rubrene crystallites. This aligned growth scheme explains why there exists an X-ray intensity correlation between rubrene and CdA films. Previous observations of enhanced nucleation 3964

dx.doi.org/10.1021/la3046912 | Langmuir 2013, 29, 3957−3967

Langmuir

Article

Figure 8. Tapping-mode AFM images (1 μm × 1 μm) for the 5 nm-thick rubrene films grown on 5L CdA layer (top two images) and on OTS/SiO2 (bottom two images). Height images are displayed on the left and the corresponding phase images on the right. Rubrene grains, appearing half dark and half bright in the phase images, are clearly discernible due to the enhanced contrast associated with the grain edges. The rubrene grains marked by arrows are those neighbored by the identifiable CdA voids, suggesting that rubrene grains germinate in the CdA voids.

are not necessarily at the same height. This misfit in vertical direction prohibits an effective π−π overlap among neighboring rubrene molecules, leading to a degraded mobility. Suffice it to say that a more diffractive CdA film is capable of producing a more crystalline rubrene film but may not guarantee a better OFET mobility because of increasing roughness at the interface between CdA dielectric and rubrene active layer.

evidence is presented to show that the sidewall interaction between rubrene and CdA is more critical than the surface energy of CdA in affecting the conformal crystalline growth of rubrene. The sidewall interaction is presumably driven energetically by a near match in molecular dimension (discrepancy